Incoming requests:
0 RPS
Static Partitioning
Training partition
8 GPUs
no sharing
Inference partition
8 GPUs
7
GPUs idle
0
Queries dropped/s
SLO violation: inference demand exceeds 8-GPU partition. Queries are being dropped.
SkyPilot Unified
Unified pool
16 GPUs
0
GPUs idle
0
Queries dropped/s
All queries served. Training yields GPUs as needed.